Very amateur notes on the UV Can Lily personal far UV-C device

•

As an interested amateur looking to learn about personal far UV-C as an infection prevention tool, I did some playing with the 222nm irradiance data from the UV Can Lily from @joeyfox85. For those interested in learning along with me, here are notes: twitter.com/joeyfox85/status/1575882717107224576

Caveat #1. Did I mention I'm an amateur? While I have earned my place to claim a fair amount infectious disease transmission expertise, when it comes to UV sterilization, I'm currently at the "white guy with a physics degree" stage of personal development. HERE BE DRAGONS.

Caveat #2. @joeyfox85 has nothing to do with me eyeballing a piece of paper he took notes on and then running away with it. Anything dumb is 100% on me, and I'm very grateful he shared the image for people like me.

If you're still with me, knowing that I'm an amateur trying out some open learning with a few folks I've recently crossed paths with on Twitter, read on.

Everything in this thread is supported with a script and the comments within here: github.com/famulare/far-UVC-exploration. Eyeballed data, a simple model of the irradiance field, and code to make some plots is here github.com/famulare/far-UVC-exploration/blob/main/uv_can_lily_CRUDE_SKETCH.R.

Here's the Lily. uv-can.com/products/lily-handheld-personal-far-uv-disinfection-light.

First thing that became really obvious is the simplest model of the light field, 1/r^2 decay from a point source, is quite wrong near the device. The axial intensity closely follows the geometry expected for the electric field of a disk with finite radius. physics.udel.edu/~watson/phys208/exercises/kevan/efield1.html

This decay is only linear out to an effective disk radius I find to be ~6.2 cm. This parameter must be a function of the reflector + lamp geometry, but I haven't tried to work out how. 1/r^2 doesn't fully kick in until >~ 20 cm axially away from the device.

Other good news is the fitted irradiance at the device face was 506 mu-W/cm^2 (and the residual around there is ~ +-15 mu-W/cm^2), which matches the manufacturer's stated value of of 490+ mu-W/cm^2. Good on UV Can for accurate specs.

Off-axis, the irradiance field >=10 cm out fits decently to the same distance dependence and falls off like cos(angle). Near the device, where there are no measurements in the tweet, we don't really know, but I can interpolate between observations and the surface intensity.

All together, we get a curve fitting model inspired by some some very basic physics for the irradiance field @joeyfox85 measured. It looks like this github.com/famulare/far-UVC-exploration/blob/f0df6fcd1125a4e7ec2f4a6dd582a6ade9a6510b/uv_can_lily_CRUDE_SKETCH.R#L108, with color is log10(irradiance) (cutoff at <1) and contours at 3,10,30,100. Looks okay! Residuals too!

(Well, the residuals are worse near the device, where the disk model is wrongest, but all are relatively small. There must be fat old books with formulae for incoherent beam cones, but I'm missing the right language and so I couldn't find them. If you have pointers, let me know.)

With the irradiance field, we can estimate that exposure time field for various amounts of viral killing. Figs below are based on pubmed.ncbi.nlm.nih.gov/35458414/, where they find one e-folding (68% reduction) takes 1/12.4*1000 = 80 mu-J/cm^2, and 10-fold reduction 80*2.3 = 180 mu-J/cm^2.

Note that this is a low required dose estimate compared to earlier work and for 254 nm light (iuva.org/resources/covid-19/Far%20UV-C%20Radiation-%20Current%20State-of%20Knowledge.pdf), but I'm using it because the Welch et al data is the most convincing to me and relevant IMO. Note also that it's a lot harder to get past 99% reductions.

Anyway, here are the estimated killing times for particles in the UV Can Lily irradiance field. So what does this mean to me?

First is that the numbers in my old thread are too pessimistic, because the pure 1/r^2 assumption isn't right. twitter.com/famulare_mike/status/1573514122075181056 I think it's no longer obvious to me that personal devices at this power rating or a bit above can't work.

Second, irradiance fields alone do not answer the question of does personal UV-C work. Virus doesn't park itself and wait to die. What matters is the dwell time of virus in the vicinity of the light source, integrated over the irradiance field, before that air gets in your face.

All that physics and biology is a lot more complicated than can be grokked solely from a light field (at the irradiance levels considered). Two great reviews: cambridge.org/core/journals/journal-of-fluid-mechanics/article/effects-of-ventilation-on-the-indoor-spread-of-covid19/CF272DAD7C27DC44F6A9393B0519CAE3 and science.org/doi/10.1126/science.abd9149.

Effectiveness will vary widely, depending on how air mixes in and into the space between the incoming air, the light, and your face. I really want to see good physics-based modeling from people selling these devices, and I'm disappointed it doesn't seem to be available yet.

Third, all that having been said, the volume of a typical breath is ~0.5 L = 500 cm^3 = a cube ~8 cm on a side, and inhalation takes a few seconds. So if I have a Lily hanging under my chin and aimed at my nose, I expect it can kill about half of the virus about to enter my face.

That's not gonna help if the person sitting across from me is spraying me down, but, for transmission from the ambient air, transmission rates are linear in viral concentration. And so a 50% cut would be nice when a mask isn't an option! And more power--safely!--would be better.

Fourth, UV safety is also a major concern. I want to see much more clarity from device sellers, and eventual regulation, as consumers like me cannot easily verify safety. twitter.com/joeyfox85/status/1575882726225301504 Leaky lights can cause harm. (interesting paper: scienceopen.com/document_file/87fb0538-10bb-4e5d-9225-9f6844d83c1f/PubMedCentral/87fb0538-10bb-4e5d-9225-9f6844d83c1f.pdf)

Fifth, I cannot imagine any circumstances where a personal UV-C of this power is reliably as effective as a well-fitting N-95-equivalent or better. The wonderful thing about a good mask is it always does the same thing. Most external conditions don't change the risk reduction.

But we can't always mask or don't always want to. And so as part of a layered strategy for harm reduction, I remain interested in the role of far UV-C in general and personal UV-C in particular. Let me know what you think, as we take this learning journey together.

You can read the unrolled version of this thread here: typefully.com/famulare_mike/DbPnT3k